is there auto load balance between 2 active webi services?

[Former Member]

Hi ,

I have not seen any kind documentation from SAP whether BOBJ can auto load balance requests in round robin fashion between 2 or more webi services.

The only theory or belief that i see & hear is that upon reaching the max connection threshold  of one webi service, bobj will start diverting the next requests to second webi service.

So now i am looking for some kind of clarification on whether auto load balance happens between multiple webi services in round robin fashion or not.

Please share if any one has idea about this.

Thanks,

Srikar

[denis_konovalov · Accepted solution]

It is not Round Robin. There are more things involved. (number of requests, their duration etc...)

The only thing is clear -- if you have 2 Webi processing servers, over a long period of time, both will have around same number of requests, will be utilized equaly.

In the short time periods, like you turn your system on and check metrics after one hour -- you might see one of the servers with more requests.

So, if reviewing monthly metrics you see one of the WPS's over utilized -- you should get SAP to investigate.

[Former Member · Accepted solution]

Thanks to everyone who tried to respond to my question.

But i just got below response from SAP:

=====================================================

I would like to inform you that CMS distributes the load (jobs) to

servers according to the server availability (maximum jobs allocated

to individual server). This is a generally how load is balanced

between CMS and various job servers and processing servers.

=====================================================

Now i am more confused. I personally did some little testing by creating 2 webi services and tried to view multiple webi reports. What i observed is  it does pick up different webi services only while at the start of report viewing. But once it picks up perticualr webi service, that same user session is sticking to previously picked up webi service and always  goes to the same process for processing.

But now SAP response is bit more confusing.

Thanks,

Srikar

[denis_konovalov · Accepted solution]

ince you open report - the session goes to one webi server and stays there no matter what you do with that report.
once you close report and open new one - the session might go to the same server or to another available.

Why do you need to know exact mechanics ?
Its not configurable or controllable.

The only indicator of load balance working properly is server metrics and number of requests processed.

[Former Member · Accepted solution]

We are in the landscape design/sizing phase and based on this its going to be significant change in design. Thanks.

[denis_konovalov · Accepted solution]

how does this affects sizing/desing ?

[Former Member · Accepted solution]

It does Denis.

If its strictly distributes load only after reaching certain threshold point like max connections say for example. then creating multiple webi services on the same host doesn't make sense. From the same angle, i added 2 seperate hosts each running one webi service. But now knowing it does some type of load balancing i would reduce those 2 hosts into 1 and create multiple webi on each host and let BOBJ handle the load distribution. Plus they are going to run on vm's

And this only webi side story. What about DSL, MDAS. So trying to  know details before customer spends $.

Thanks.

[denis_konovalov · Accepted solution]

did you look here :

it has best practices for webi sizing/distribution as well as DSL, MDAS etc...

[Former Member]

its not round robin basis.

[ragove]

I agree to what Arvind said. I have practically experienced that request goes to any server depending upon their availability- hence, its not round robin.

[Former Member]

Hi,

Load balancing does happen between all BO services in cluster. The algorithm or the logic with which it happens is unknown.

Regards

Chinmaya

[Former Member]

This is a really good question for a nice discussion and I would also like to understand everyone's thoughts on this question.

Logically speaking I don't know how a database can assign or distribute requests in a particular manner. I believe that the requests would be getting passed randomly to any service thats available at any point of time.

Can CMS judge the service that would accept requests? If yes how?